AI's Double Edge: Will ChatGPT Replace Jobs or Transform Them?

Since late November, when ChatGPT launched and began spreading across social media, Google searches about artificial intelligence have spiked across Brazil—and nowhere more than in the Federal District, where residents are searching harder than anywhere else in the country for answers to basic questions: What is ChatGPT? How does AI work? Can software write for me?

The timing is no accident. ChatGPT, a chatbot built by OpenAI, became an overnight phenomenon because it does something that felt impossible until it happened: it writes coherent, detailed text on almost any subject in seconds. The tool draws from existing internet content and transforms user questions into responses that are often creative, sometimes eerily precise. But its arrival has triggered a deeper anxiety. Will this replace jobs, or transform them?

Marcos Barreto, a professor at Fundação Vanzolini and the Polytechnic School at USP, sees both. Any profession built on text production—writing, coding, legal briefs, medical notes—can be touched by this technology. Even doctors could benefit: ChatGPT could draft clearer prescriptions with better dosage instructions. But Barreto is careful to note what history shows: every technological leap reduces the number of people needed for certain tasks. The factory worker didn't disappear when robots arrived; the worker simply did different work. With ChatGPT, he predicts, we'll need people to review and validate what the machine produces. The job changes shape; it doesn't vanish.

Hélio Zylberstajn, a senior professor of economics at USP and coordinator of the Salariômetro project, sees something sharper. Until now, automation has replaced routine, non-cognitive work—the kind that doesn't require thinking. ChatGPT changes that equation. Lawyers, teachers, academics—people whose work demands reasoning and judgment—now face a formidable competitor. Like all technological shifts, Zylberstajn says, new opportunities will emerge. But existing ones will be destroyed. The question is which ones, and how fast.

Leon Sólon, a data science professor, offers a more granular view. ChatGPT works best as a productivity multiplier, like a calculator for thinking. A programmer who understands code can use it to speed up work that would take hours; someone who doesn't understand code might not catch the errors the tool introduces. The platform has built-in feedback mechanisms—thumbs up, thumbs down—so users can signal when responses are wrong, feeding that human judgment back into the system's training. But Sólon identifies translation as particularly vulnerable. ChatGPT can translate fluently, making text more coherent, adjusting tone and formality. Yet it struggles with specialized knowledge. A human translator who specializes in vegan philosophy, for instance, will catch nuances and terminology that ChatGPT renders generically. Expertise in a narrow domain still has armor.

In Brasília, where the federal government and a growing tech sector create one of Brazil's largest markets for digital services, Marco Tulio, president of the IT Services Union for the Federal District, takes an optimistic view—but not a naive one. The technology will evolve and integrate into nearly every system humans interact with. New companies and services will be born from it. But Tulio is clear-eyed about the risks. ChatGPT is not yet ready to assume human jobs wholesale. These systems can make mistakes. They can be programmed with hidden agendas—government policies, extremist ideologies, sales targets. What will definitely change is repetitive work. Tasks that demand hours of manual computer labor will be done in seconds. The market will shift.

Matheus Duarte, a 22-year-old product design student at the Federal Institute of Brasília, sees ChatGPT as a tool, not a threat. He uses it to understand what content works for social media, to help organize posts for the orthopedic clinic where he interns. It's faster than Google, more specific. He's not worried about his profession disappearing; he's using the tool to do his work better. Leonardo Tomazeli Duarte, a professor at Unicamp and scientific director of the Brazilian Institute of Data Science, echoes this view but with longer perspective. In the short term, ChatGPT will change how people work—how they search, synthesize information, draft first versions. In the long term, no one can say. These tools depend entirely on content already available online; they don't create original knowledge. Their limits remain unknown. What we can see now is a tool of support, not replacement. But the next decades will tell the real story.